Digital images are often annotated to support the use of browsing, searching, and retrieving features of an image retrieval system. For example, an image retrieval system may include a database of digital images having associated annotations. The annotations may include metadata such as captioning, keywords, or other descriptions associated with the digital images so that retrieval can be executed based on the annotations. In some examples, an image retrieval system may offer an automatic image annotation component.
Automatic image annotation or automatic image tagging offers the ability to automatically assign metadata to a digital image. In some instances, computer vision techniques are employed to organize or locate images of interest from a database. For example, a system may include training images having annotations associated with various feature vectors (classifications of colors, shapes, textures, patterns, etc. within an image). Machine learning may then be used to determine a correlation between the feature vectors and the associated annotations of the training images. Finally, annotations may be assigned to raw or un-annotated images based on feature vectors identified within the raw images and the correlations determined via machine learning.